A recycling and identification method for waste lithium batteries based on image analysis

Abstract

The present invention discloses a new energy waste lithium battery recycling and identification method based on image analysis, which relates to the field of image analysis and identification technology, including collecting lithium battery images through installation equipment and preprocessing; extracting feature graphs of the processed lithium battery images, using a convolutional neural network (CNN) model for image recognition, classifying and processing the recognition results, evaluating the classification results and making adjustments; and storing all data. The present invention uses the Sobel operator and the Canny edge detection algorithm to perform accurate edge detection on the processed images, and uses the convolutional neural network (CNN) to perform in-depth feature learning and identification on lithium battery images, which significantly improves the accuracy and efficiency of waste lithium battery identification and classification, while reducing dependence on manual operations, and provides a safer and more reliable technical solution for the efficient recycling of waste lithium batteries.

Description

Technical Field

[0001] The invention relates to the technical field of image analysis and recognition, in particular to a new energy waste lithium battery recycling and recognition method based on image analysis. Background Art

[0002] In the current context of environmental protection and resource recycling, the recycling and reuse of new energy batteries, especially lithium batteries, has become an important issue in environmental protection and resource conservation. Lithium batteries are widely used in various electronic devices and electric vehicles due to their excellent energy density and long service life. Improper handling of used lithium batteries will cause serious pollution to the environment and waste the precious metals and other chemicals contained in the batteries. It is particularly important to develop efficient and accurate lithium battery recycling and identification technology.

[0003] Existing lithium battery recycling technologies mainly rely on manual sorting and traditional mechanical processing methods. These methods are not only inefficient, but also have problems with inaccurate identification, which can easily lead to the loss of reusable resources and increased environmental pollution risks. Especially when there are many types of batteries with different shapes, traditional methods are difficult to quickly and accurately identify and classify, which limits the recycling efficiency and quality. The judgment of the physical state of the battery often depends on the experience of the operator and is easily affected by subjective factors, increasing operational risks. Summary of the invention

[0004] In view of the problems existing in the above-mentioned existing new energy waste lithium battery recycling and identification method based on image analysis, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is that the existing lithium battery recycling technology mainly relies on manual sorting and traditional mechanical processing methods. These methods are not only inefficient, but also have the problem of inaccurate identification, which easily leads to the loss of reusable resources and the increase of environmental pollution risks. Traditional methods are difficult to achieve fast and accurate identification and classification, which limits the recycling efficiency and quality.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for recycling and identifying waste lithium batteries for new energy based on image analysis, which includes collecting lithium battery images by installing equipment and preprocessing them; extracting feature maps of the processed lithium battery images, using a convolutional neural network (CNN) model for image recognition, classifying and processing the recognition results, evaluating the classification results and making adjustments; and storing all data.

[0007] As a preferred solution of the new energy waste lithium battery recycling and identification method based on image analysis of the present invention, wherein: the acquisition of lithium battery images by installing equipment refers to fixing a high-resolution industrial camera above the conveyor belt, and an annular LED light source is arranged around the camera;

[0008] A photoelectric sensor is installed on the conveyor belt. When the battery passes through the sensor, the camera is triggered to take images of the waste lithium batteries.

[0009] As a preferred solution of the new energy waste lithium battery recycling identification method based on image analysis of the present invention, the preprocessing refers to using a Gaussian filtering algorithm to remove noise in the image and setting the standard deviation of the Gaussian kernel. is 1.5, the formula is:

[0010]

[0011] Where A(u,v) is the filtered image The pixel value of the point, u and v are the pixel positions of the image;

[0012] Adaptive histogram equalization (CLAHE) technology is used to enhance the contrast of the denoised image.

[0013] The LAHE processed image C is converted into a grayscale image using the formula:

[0014]

[0015] Where E is the grayscale value, R, G and B are the values ​​of the red, green and blue channels;

[0016] Combine the grayscale values ​​of the pixels into a grayscale image Y;

[0017] The content of using adaptive histogram equalization CLAHE technology is:

[0018] Split the image into 8x8 grid blocks;

[0019] The histogram equalization formula is:

[0020]

[0021] in is the pixel value of the image after equalization, is the pixel value of the image after filtering, and are the minimum and maximum pixel values ​​of the image;

[0022] The equalized small blocks are reassembled into image C.

[0023] As a preferred solution of the new energy waste lithium battery recycling and identification method based on image analysis of the present invention, the feature map of the extracted lithium battery image refers to convolution operation of the Sobel kernel matrix with the grayscale image Y, and the gradient in the x and y directions is calculated using the Sobel operator, and the formula is:

[0024]

[0025]

[0026] in and is the gradient of the image in the horizontal and vertical directions;

[0027] Use what you get and Calculate the gradient M and gradient direction of the pixel , the formula is:

[0028]

[0029]

[0030] Use the Canny edge detection algorithm to extract the edges of the battery:

[0031] Check along the gradient direction at the pixel point. If the gradient amplitude of the current pixel is not the local maximum, suppress it to 0.

[0032] Set high threshold and low threshold , for gradient M above a high threshold The pixel point is marked as a strong edge pixel point, and the gradient M is lower than the low threshold The pixel points are suppressed as non-edge pixels, and the gradient M is between the high threshold and low threshold If the pixels between are connected to a strong edge, they are marked as edge pixels, otherwise they are suppressed as non-edge pixels;

[0033] The edge shape features of the image are obtained, including the contour area S and the contour perimeter P. The formula is:

[0034]

[0035]

[0036] Where m is the number of points of the contour, is the coordinate of the i-th point on the contour;

[0037] The gradient amplitude M and gradient direction of the pixel point , contour area S and contour perimeter P are integrated into the feature map, the formula is:

[0038]

[0039] Where F is the feature map.

[0040] As a preferred solution of the new energy waste lithium battery recycling and identification method based on image analysis described in the present invention, wherein: the use of the convolutional neural network (CNN) model for image recognition refers to standardizing the historical image data of waste lithium batteries to generate a historical training set and annotating them into two categories: recyclable and non-recyclable;

[0041] The definition of the convolutional neural network CNN model includes input layer, convolution layer, activation function, pooling layer, fully connected layer, and output layer:

[0042] Define the input layer, the input is the normalized image;

[0043] Use the convolutional layer to extract image features. The formula is:

[0044]

[0045] Where O is the convolution output, I is the normalized image, K is the convolution kernel, b is the bias term, and k is the convolution kernel size;

[0046] Using the ReLU activation function, the formula is:

[0047]

[0048] in is the input value;

[0049] Use maximum pooling to reduce the size of the feature map. The formula is:

[0050]

[0051] Where P is the pooling output, z and r are the pooling window sizes;

[0052] Flatten the pooling layer output into a one-dimensional vector , connected to the fully connected layer, the formula is:

[0053]

[0054] in is the output of the zth neuron, is the weight vector of the zth neuron, is the flattened input feature vector, The bias scalar of the zth neuron;

[0055] Use the Softmax function for classification and output the probability of each category. The formula is:

[0056]

[0057] Where W is the total number of categories;

[0058] Use the training set to train the convolutional neural network (CNN) model, and use the cross entropy loss function and Adam optimizer for iterative parameter optimization;

[0059] Use the convolutional neural network CNN model to output the probability of the category, determine the category of the image based on the highest probability value, set the output type to N, and the final recognition result is:

[0060]

[0061] The feature map F is normalized and input into the trained convolutional neural network (CNN) model to obtain the image recognition result N.

[0062] As a preferred solution of the new energy waste lithium battery recycling identification method based on image analysis of the present invention, wherein: the classification and processing of lithium batteries according to the identification results refers to classifying lithium batteries into recyclable and non-recyclable categories according to the output result N of the convolutional neural network CNN model, and sorting the lithium batteries to the corresponding processing area by an automatic sorting device:

[0063] If N is classified as a recyclable lithium battery, the recyclable lithium battery is transported to the recycling processing area for subsequent processing;

[0064] If N is classified as non-recyclable lithium battery, the non-recyclable battery is transported to the waste treatment area for safe disposal.

[0065] As a preferred solution of the new energy waste lithium battery recycling and identification method based on image analysis of the present invention, wherein: the evaluation of the classification results and adjustment refers to randomly selecting the image recognition results N for manual re-inspection and marking;

[0066] The accuracy of the output result N of the convolutional neural network CNN model is calculated using the formula:

[0067]

[0068] Where A is the accuracy rate, TP is the number of lithium batteries correctly classified as recyclable, TN is the number of lithium batteries correctly classified as non-recyclable, FP is the number of non-recyclable lithium batteries incorrectly classified as recyclable, and FN is the number of recyclable lithium batteries incorrectly classified as non-recyclable;

[0069] Adjust the convolution kernel size and pooling kernel size model parameters according to the evaluation results.

[0070] As a preferred solution of the new energy waste lithium battery recycling and identification method based on image analysis described in the present invention, wherein: the storing of all data refers to storing images collected by the device, denoised images, image data with contrast enhanced by adaptive histogram equalization CLAHE technology, extracted feature data, label data for battery classification, convolutional neural network CNN model output data, classification processing data and evaluation adjustment data in a database.

[0071] A computer device comprises: a memory and a processor; the memory stores a computer program, and the processor implements the steps of the above-mentioned new energy waste lithium battery recycling and identification method based on image analysis when executing the computer program.

[0072] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned new energy waste lithium battery recycling and identification method based on image analysis.

[0073] The beneficial effects of the present invention are as follows: the present invention uses the Sobel operator and the Canny edge detection algorithm to perform accurate edge detection on the processed image, and uses the convolutional neural network CNN to perform in-depth feature learning and recognition of the lithium battery image, thereby significantly improving the accuracy and efficiency of identifying and classifying waste lithium batteries, while reducing dependence on manual operations, and providing a safer and more reliable technical solution for the efficient recycling of waste lithium batteries. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0075] Figure 1 The figure is a flow chart of the new energy waste lithium battery recycling and identification method based on image analysis.

[0076] Figure 2 This is a structural diagram of the new energy waste lithium battery recycling and identification method based on image analysis. DETAILED DESCRIPTION

[0077] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0078] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0079] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0080] Example 1

[0081] Reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a new energy waste lithium battery recycling identification method based on image analysis. The new energy waste lithium battery recycling identification method based on image analysis includes:

[0082] S1. Collect lithium battery images by installing equipment and perform preprocessing;

[0083] Specifically, collecting lithium battery images by installing equipment means fixing a high-resolution industrial camera above a conveyor belt, and arranging a ring-shaped LED light source around the camera;

[0084] A photoelectric sensor is installed on the conveyor belt. When the battery passes through the sensor, the camera is triggered to take images of the waste lithium batteries.

[0085] The detailed images captured by high-resolution industrial cameras, combined with the uniform illumination provided by the ring-shaped LED light source, greatly improve the image quality. This enables the image analysis algorithm to more accurately identify the type of battery, degree of damage and other key features, such as the battery model, size and possible defects. The introduction of photoelectric sensors automates the identification process and reduces the need for manual intervention, thereby reducing labor costs and increasing processing speed. The automated process not only improves the efficiency of the overall operation, but also increases the feasibility of processing large quantities of waste batteries.

[0086] Furthermore, preprocessing refers to using a Gaussian filter algorithm to remove noise from the image and setting the standard deviation of the Gaussian kernel. is 1.5, the formula is:

[0087]

[0088] Where A(u,v) is the filtered image The pixel value of the point, u and v are the pixel positions of the image;

[0089] Adaptive histogram equalization (CLAHE) technology is used to enhance the contrast of the denoised image.

[0090] The LAHE processed image C is converted into a grayscale image using the formula:

[0091]

[0092] Where E is the grayscale value, R, G and B are the values ​​of the red, green and blue channels;

[0093] Combine the grayscale values ​​of the pixels into a grayscale image Y;

[0094] The content of using adaptive histogram equalization CLAHE technology is:

[0095] Split the image into 8x8 grid blocks;

[0096] The histogram equalization formula is:

[0097]

[0098] in is the pixel value of the image after equalization, is the pixel value of the image after filtering, and are the minimum and maximum pixel values ​​of the image;

[0099] The equalized small blocks are reassembled into image C.

[0100] The application of Gaussian filtering and CLAHE technology can significantly improve the quality of waste lithium battery images. Clear images are the basis of high-precision image analysis, ensuring higher recognition rates and classification accuracy. Clear images with appropriate contrast make it easier to identify and analyze the characteristics of lithium batteries, such as model and damage status, through image processing algorithms, thereby improving the automation level and processing efficiency of the entire system. Compared with traditional manual detection or mechanical screening methods, the present invention reduces labor costs and speeds up processing through automated image analysis, and is particularly suitable for processing large-scale waste batteries, with significant economic benefits.

[0101] S2, extracting the feature map of the processed lithium battery image, using the convolutional neural network (CNN) model for image recognition, classifying and processing the recognition results, evaluating the classification results and making adjustments;

[0102] Specifically, extracting the feature map of the processed lithium battery image means performing a convolution operation on the Sobel kernel matrix and the grayscale image Y, and using the Sobel operator to calculate the gradients in the x and y directions, and the formula is:

[0103]

[0104]

[0105] in and is the gradient of the image in the horizontal and vertical directions;

[0106] Use what you get and Calculate the gradient M and gradient direction of the pixel , the formula is:

[0107]

[0108]

[0109] Use the Canny edge detection algorithm to extract the edges of the battery:

[0110] Check along the gradient direction at the pixel point. If the gradient amplitude of the current pixel is not the local maximum, suppress it to 0.

[0111] Set high threshold and low threshold , for gradient M above the high threshold The pixel point is marked as a strong edge pixel point, and the gradient M is lower than the low threshold The pixel points are suppressed as non-edge pixels, and the gradient M is between the high threshold and low threshold If the pixels between are connected to a strong edge, they are marked as edge pixels, otherwise they are suppressed as non-edge pixels;

[0112] The edge shape features of the image are obtained, including the contour area S and the contour perimeter P. The formula is:

[0113]

[0114]

[0115] Where m is the number of points of the contour, is the coordinate of the i-th point on the contour;

[0116] The gradient amplitude M and gradient direction of the pixel point , contour area S and contour perimeter P are integrated into the feature map, the formula is:

[0117]

[0118] Where F is the feature map.

[0119] By using the Sobel operator and the Canny edge detection algorithm, the present invention can effectively identify the precise edges and contours of the battery, which is crucial for determining the type and state of the battery. Accurate contour information helps in subsequent classification and processing decisions, improving the efficiency and accuracy of the entire system. Automated image processing reduces dependence on manual operations, significantly improves processing speed and reduces labor costs, which is particularly important for large-scale battery recycling operations. It can process larger quantities of batteries and accelerate the entire recycling process. Through efficient and accurate classification, recyclable batteries can be effectively recycled, while damaged or harmful batteries are properly handled, reducing environmental pollution. Resource recycling also helps to reduce the demand for virgin materials, in line with the goal of sustainable development. The extracted feature map F contains information such as the gradient amplitude, gradient direction, contour area and perimeter of the battery. These data provide rich features for machine learning algorithms and can be used to train more accurate classification models, such as convolutional neural networks.

[0120] Furthermore, using a convolutional neural network (CNN) model for image recognition means standardizing the historical image data of waste lithium batteries to generate a historical training set and annotating them into two categories: recyclable and non-recyclable;

[0121] The definition of the convolutional neural network CNN model includes input layer, convolution layer, activation function, pooling layer, fully connected layer, and output layer:

[0122] Define the input layer, the input is the normalized image;

[0123] Use the convolutional layer to extract image features. The formula is:

[0124]

[0125] Where O is the convolution output, I is the normalized image, K is the convolution kernel, b is the bias term, and k is the convolution kernel size;

[0126] The feature map F mainly calculates the gradient amplitude and direction through the Sobel operator, emphasizing the edge and contour features of the image. The convolution layer performs convolution operations through multiple convolution kernels to extract multi-level local features, covering feature information from low-level to high-level;

[0127] Using the ReLU activation function, the formula is:

[0128]

[0129] in is the input value;

[0130] Use maximum pooling to reduce the size of the feature map. The formula is:

[0131]

[0132] Where P is the pooling output, z and r are the pooling window sizes;

[0133] Flatten the pooling layer output into a one-dimensional vector , connected to the fully connected layer, the formula is:

[0134]

[0135] in is the output of the zth neuron, is the weight vector of the zth neuron, is the flattened input feature vector, The bias scalar of the zth neuron;

[0136] Use the Softmax function for classification and output the probability of each category. The formula is:

[0137]

[0138] Where W is the total number of categories;

[0139] Use the training set to train the convolutional neural network (CNN) model, and use the cross entropy loss function and Adam optimizer for iterative parameter optimization;

[0140] Use the convolutional neural network CNN model to output the probability of the category, determine the category of the image based on the highest probability value, set the output type to N, and the final recognition result is:

[0141]

[0142] The feature map F is normalized and input into the trained convolutional neural network (CNN) model to obtain the image recognition result N.

[0143] Through the fine-tuned CNN model, recyclable and non-recyclable lithium batteries can be accurately distinguished. This classification is based on the visual features of the battery, such as the degree of damage, shape and markings. These features are extracted through advanced image processing technology and used to train the model. The automated image recognition system has greatly improved the speed and magnitude of waste battery processing. Compared with the traditional manual sorting method, it greatly reduces costs and time consumption, while reducing human errors. Effective classification and recycling of waste lithium batteries is of great significance to environmental protection. By reducing misclassification, ensuring that hazardous substances are properly handled, and maximizing the recovery of valuable materials, the application of the CNN model provides the possibility of continuous improvement and optimization. By continuously learning new data, the model can adapt to new battery types and emerging recycling standards to ensure the long-term effectiveness and adaptability of the system.

[0144] Furthermore, the identification results are classified and processed into lithium batteries, which means that the lithium batteries are classified into recyclable and non-recyclable categories according to the output result N of the convolutional neural network CNN model, and the lithium batteries are sorted to the corresponding processing area through an automatic sorting device:

[0145] If N is classified as a recyclable lithium battery, the recyclable lithium battery is transported to the recycling processing area for subsequent processing;

[0146] If N is classified as non-recyclable lithium battery, the non-recyclable battery is transported to the waste treatment area for safe disposal.

[0147] The use of CNN models can significantly improve the accuracy of lithium battery classification. The model can learn the subtle differences in battery images, including signs of damage and degree of wear, ensuring that each battery is accurately classified, thereby optimizing the recycling process and resource utilization. The automatic sorting device automatically operates according to the judgment results of the CNN model. It is faster and more efficient than traditional manual sorting, can handle a large number of batteries, and adapt to high-load working environments. Accurately classifying batteries and handling them properly can reduce environmental pollution, especially for damaged or potential safety hazards. Batteries ensure that they are safely handled and prevent harmful substances from leaking into the environment. Through an efficient classification system, recyclable batteries are properly recycled and processed, maximizing the recovery of precious metals and chemicals, reducing the demand and mining of raw materials, and the automatic system reduces manual operations, especially when handling potentially harmful waste batteries, reducing health risks for workers.

[0148] Specifically, evaluating the classification results and making adjustments means randomly selecting N image recognition results for manual re-examination and labeling;

[0149] The accuracy of the output result N of the convolutional neural network CNN model is calculated using the formula:

[0150]

[0151] Where A is the accuracy rate, TP is the number of lithium batteries correctly classified as recyclable, TN is the number of lithium batteries correctly classified as non-recyclable, FP is the number of non-recyclable lithium batteries incorrectly classified as recyclable, and FN is the number of recyclable lithium batteries incorrectly classified as non-recyclable;

[0152] Adjust the convolution kernel size and pooling kernel size model parameters according to the evaluation results. If the accuracy is low, you need to reduce the learning rate or increase the number of training rounds.

[0153] By introducing a manual re-inspection and labeling mechanism, the present invention can not only correct preliminary classification errors, but also optimize the training of the CNN model through these data, so that the model can more accurately adapt to changes and characteristics in practical applications. By dynamically adjusting model parameters such as the size of convolution kernels and pooling kernels, the present invention can adjust the model for different lithium battery types and damage conditions, improve its generalization ability and flexibility, and more accurate classification results reduce the misjudgment of resources, ensure the effective utilization of recyclable materials, and avoid the erroneous processing of non-recyclable materials. The mechanism of automation combined with manual re-inspection improves the processing speed, greatly reduces the time cost compared to full manual inspection, and improves the efficiency of processing waste lithium batteries. By improving the accuracy and efficiency of lithium battery recycling, the present invention supports the sustainable utilization of resources and environmental protection, which is in line with global sustainable development goals.

[0154] S3, store all data;

[0155] Furthermore, storing all data means storing images collected by the device, denoised images, image data with contrast enhanced by adaptive histogram equalization CLAHE technology, extracted feature data, label data for battery classification, convolutional neural network CNN model output data, classification processing data, and evaluation adjustment data in a database.

[0156] By centrally storing all relevant data, the system is able to quickly access and process necessary information, thereby providing data to support decisions when needed, such as adjusting classification parameters in real time or retraining machine learning models. The integrated stored data can be used to generate in-depth analysis reports to help monitor the efficiency and effectiveness of the entire recycling process, thereby achieving continuous improvement of the process. Through precise classification and efficient recycling processes, this system helps to reduce environmental pollution, improve resource utilization, and support the goals of environmental protection and sustainable development.

[0157] Example 2

[0158] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0159] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0160] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0161] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

Claims

1. A new energy waste lithium battery recycling and identification method based on image analysis, characterized in that: include, Collect lithium battery images and pre-process them by installing equipment; Extract the feature map of the processed lithium battery image, use the convolutional neural network (CNN) model for image recognition, classify and process the recognition results, evaluate the classification results and make adjustments; Store all data; The collecting of lithium battery images by installing equipment refers to fixing a high-resolution industrial camera above a conveyor belt, and arranging a ring-shaped LED light source around the camera; Install a photoelectric sensor on the conveyor belt. When the battery passes the sensor, the camera is triggered to take pictures of the waste lithium battery. The preprocessing refers to using a Gaussian filtering algorithm to remove noise in the image, setting the standard deviation σ of the Gaussian kernel to 1.5, and the formula is: Where A(u,v) is the pixel value of the filtered image point (u,v), and u and v are the pixel positions of the image; Adaptive histogram equalization (CLAHE) technology is used to enhance the contrast of the denoised image. The LAHE processed image C is converted into a grayscale image using the formula: E=0.299R+0.587G+0.114B Where E is the grayscale value, R, G and B are the values ​​of the red, green and blue channels; Combine the grayscale values ​​of the pixels into a grayscale image Y; The content of using adaptive histogram equalization CLAHE technology is: Split the image into 8x8 grid blocks; The histogram equalization formula is: Where C(G) is the pixel value of the image after equalization, G is the pixel value of the image after filtering, and G min and G max are the minimum and maximum pixel values ​​of the image; Reassemble the equalized small blocks into image C; The feature map of the extracted lithium battery image refers to the convolution operation of the Sobel kernel matrix with the grayscale image Y, and the Sobel operator is used to calculate the gradients in the x and y directions, and the formula is: Where S x and S y is the gradient of the image in the horizontal and vertical directions; Use the obtained S x and S y Calculate the gradient M and gradient direction θ of the pixel point, the formula is: Use the Canny edge detection algorithm to extract the edges of the battery: Check along the gradient direction at the pixel point. If the gradient amplitude of the current pixel is not the local maximum, suppress it to 0. Set the high threshold T h and low threshold T l , for gradient M above a high threshold T h The pixel point is marked as a strong edge pixel point, and the gradient M is lower than the low threshold T l The pixel points are suppressed as non-edge pixels, and the gradient M is between the high threshold T h and low threshold T l If the pixels between are connected to a strong edge, they are marked as edge pixels, otherwise they are suppressed as non-edge pixels; The edge shape features of the image are obtained, including the contour area S and the contour perimeter P. The formula is: Where m is the number of points of the contour, (x i ,y i ) is the coordinate of the i-th point on the contour; The gradient amplitude M, gradient direction θ, contour area S and contour perimeter P of the pixel point are integrated into the feature map. The formula is: F=[M,θ,S,P] Where F is the feature map.

2. The new energy waste lithium battery recycling and identification method based on image analysis as claimed in claim 1 is characterized by: The use of a convolutional neural network (CNN) model for image recognition refers to standardizing the historical image data of waste lithium batteries to generate a historical training set and annotating them into two categories: recyclable and non-recyclable; The definition of the convolutional neural network CNN model includes input layer, convolution layer, activation function, pooling layer, fully connected layer, and output layer: Define the input layer, the input is the normalized image; Use the convolutional layer to extract image features. The formula is: Where O is the convolution output, I is the normalized image, K is the convolution kernel, b is the bias term, and k is the convolution kernel size; Using the ReLU activation function, the formula is: ReLU(τ)=max(0,τ) Where τ is the input value; Use maximum pooling to reduce the size of the feature map. The formula is: P i,j =max{O i+z,j+r } Where P is the pooling output, z and r are the pooling window sizes; Flatten the pooling layer output into a one-dimensional vector ω z , connected to the fully connected layer, the formula is: oh z =Q z ·μ+U z where ω z is the output of the zth neuron, Q z is the weight vector of the zth neuron, μ is the flattened input feature vector, U z The bias scalar of the zth neuron; Use the Softmax function for classification and output the probability of each category. The formula is: Where W is the total number of categories; Use the training set to train the convolutional neural network (CNN) model, and use the cross entropy loss function and Adam optimizer for iterative parameter optimization; Use the convolutional neural network CNN model to output the probability of the category, determine the category of the image based on the highest probability value, set the output type to N, and the final recognition result is: N=argmax i (Softmax(ω z )) The feature map F is normalized and input into the trained convolutional neural network (CNN) model to obtain the image recognition result N.

3. The new energy waste lithium battery recycling and identification method based on image analysis according to claim 2 is characterized in that: The classification and processing of lithium batteries according to the identification results refers to classifying lithium batteries into recyclable and non-recyclable categories according to the output result N of the convolutional neural network CNN model, and sorting the lithium batteries to corresponding processing areas through an automatic sorting device: If N is classified as a recyclable lithium battery, the recyclable lithium battery is transported to the recycling processing area for subsequent processing; If N is classified as non-recyclable lithium battery, the non-recyclable battery is transported to the waste treatment area for safe disposal.

4. The new energy waste lithium battery recycling and identification method based on image analysis as claimed in claim 3 is characterized by: The evaluating and adjusting the classification results refers to randomly selecting the image recognition results N for manual re-examination and marking; The accuracy of the output result N of the convolutional neural network CNN model is calculated using the formula: Where A is the accuracy rate, TP is the number of lithium batteries correctly classified as recyclable, TN is the number of lithium batteries correctly classified as non-recyclable, FP is the number of non-recyclable lithium batteries incorrectly classified as recyclable, and FN is the number of recyclable lithium batteries incorrectly classified as non-recyclable; Adjust the convolution kernel size and pooling kernel size model parameters according to the evaluation results.

5. The new energy waste lithium battery recycling and identification method based on image analysis as claimed in claim 4 is characterized in that: Storing all data refers to storing images collected by the device, denoised images, image data with contrast enhanced by adaptive histogram equalization CLAHE technology, extracted feature data, label data for battery classification, convolutional neural network CNN model output data, classification processing data and evaluation adjustment data in the database.

6. A computer device comprising: Memory and processor; The memory stores a computer program, characterized in that: when the processor executes the computer program, the steps of the new energy waste lithium battery recycling and identification method based on image analysis described in any one of claims 1 to 5 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the new energy waste lithium battery recycling and identification method based on image analysis described in any one of claims 1 to 5 are implemented.

Images